ansaurus

Question

Is there a way to know which versions of python are supported by my code?

Answer 1

+7 A:

In response to a previous question about this, I wrote pyqver. If you have any improvements, please feel free to fork and contribute!

Greg Hewgill 2009-07-13 08:09:04

+1 for awesome hair color

lyrae 2009-07-13 08:25:56

pyqver seems to be the tool I am looking for. Thank you very much.However, I've notice that my script was seen as compatible with 2.5 but it was not. from __future__ import with_statement was missing.The "except as" was also not detected as a compatibility issue with 2.5. I can imagine that it is complex to handle all cases. pyqver is a great tool and I will try to contribute and to improve it if I can

luc 2009-07-13 09:30:50

The "except as" is noted as a bug in the documentation. I'll see if I can find out what happened with the with statement.

Greg Hewgill 2009-07-13 09:37:52

When you noted the problem with the "with" statement, which version of Python were you using to run pyqver? It works as expected when run with 2.6, but with 2.5 the standard library "compiler" module doesn't recognise the "with" keyword. I recommend to run pyqver with 2.6 or later.

Greg Hewgill 2009-07-13 09:43:53

I was running pyqver with python 2.6. It detected my code as valid for python 2.5. In fact my code was valid for that version but the "from future import with_statement" was missing.

luc 2009-07-13 10:03:40

This approach may be useful for *triage* of a script/module/package that needs to be made usable on older Pythons, but there is no substitute for actual testing; many problems can't be discovered by static analysis (e.g. a in b where b is a dict). Notes on the tool itself: (1) doctest fails for spurious reasons (actual output is not in the simplistic (2, 5) format of the expected) (2) handling e.g. `from decimal import Decimal` is in the TODO

John Machin 2009-07-13 10:42:53

@John: I noticed the broken doctest too and committed a fix.

Greg Hewgill 2009-07-13 10:48:33

Answer 2

+5 A:

I recommend you rather use automated tests than a code analysis tool.

Be aware that there are subtle behaviour changes in the Python standard library that your code may or may not depend upon. For example httplib: When uploading files, it is normal to give the data as a str. In Python 2.6 you can give stream objects instead (useful for >1GB files) if you nudge them correctly, but in Python 2.5 you will get an error.

A comprehensive set of unit tests and integration tests will be much more reliable because they test that your program actually works on Python version X.Y.

$ python2.6 tests/run_all.py
.................................
33 tests passed
[OK]

You're Python 2.6 compatible.

$ python2.4 tests/run_all.py
...........EEE.........EEE.......
27 tests passed, 6 errors
[FAIL]

You're not Python 2.4 compatible.

Deestan 2009-07-13 09:24:52

+1 This is also a good idea. There are quite a range of more subtle things that a static analysis tool cannot detect. If you've got good coverage with your unit tests, this is an excellent method.

Greg Hewgill 2009-07-13 09:35:10

Thansk for this remark which is very valuable. However I don't see a code analysis tool as a competitor for unit testing. It has different purposes and should be used together. Analysis tool can help detect potential problems more quickly than unit testing because it is focussed on compatibility issues. I am very happy to get a code analysis tool in my toolbox now but of course I keep running my tests for all supported versions

luc 2009-07-13 10:09:50

Answer 3

A:

Python 2.5 can still be saved, since it can use the with keyword:

from __future__ import with_statement

kaizer.se 2009-07-13 09:51:55

ansaurus

tags:

views:

answers:

Is there a way to know which versions of python are supported by my code?

related questions